Limits...
Enhanced CellClassifier: a multi-class classification tool for microscopy images.

Misselwitz B, Strittmatter G, Periaswamy B, Schlumberger MC, Rout S, Horvath P, Kozak K, Hardt WD - BMC Bioinformatics (2010)

Bottom Line: This represents a significant obstacle in many biology laboratories.Classification results can be integrated with other object measurements including inter-object relationships.This should facilitate the implementation of automated high-content screening.

View Article: PubMed Central - HTML - PubMed

Affiliation: Institute of Microbiology, ETH Zurich, Zürich, Switzerland. misselwitz@micro.biol.ethz.ch

ABSTRACT

Background: Light microscopy is of central importance in cell biology. The recent introduction of automated high content screening has expanded this technology towards automation of experiments and performing large scale perturbation assays. Nevertheless, evaluation of microscopy data continues to be a bottleneck in many projects. Currently, among open source software, CellProfiler and its extension Analyst are widely used in automated image processing. Even though revolutionizing image analysis in current biology, some routine and many advanced tasks are either not supported or require programming skills of the researcher. This represents a significant obstacle in many biology laboratories.

Results: We have developed a tool, Enhanced CellClassifier, which circumvents this obstacle. Enhanced CellClassifier starts from images analyzed by CellProfiler, and allows multi-class classification using a Support Vector Machine algorithm. Training of objects can be done by clicking directly "on the microscopy image" in several intuitive training modes. Many routine tasks like out-of focus exclusion and well summary are also supported. Classification results can be integrated with other object measurements including inter-object relationships. This makes a detailed interpretation of the image possible, allowing the differentiation of many complex phenotypes. For the generation of the output, image, well and plate data are dynamically extracted and summarized. The output can be generated as graphs, Excel-files, images with projections of the final analysis and exported as variables.

Conclusion: Here we describe Enhanced CellClassifier which allows multiple class classification, elucidating complex phenotypes. Our tool is designed for the biologist who wants both, simple and flexible analysis of images without requiring programming skills. This should facilitate the implementation of automated high-content screening.

Show MeSH

Related in: MedlinePlus

Feature integration by CellClassifier. A: Salmonella (shown in green) preferentially dock onto mitotic cells (nuclei shown in grey). Segmentation and measurements of image were done using CellProfiler: First both, nuclei and Salmonella, were identified as independent objects. Cell objects were generated by expansion with reference to the nucleus. Inter-object relationships between Salmonella-spots and cells, as well as between neighboring nuclei were calculated. In CellClassifier a model was trained to distinguish mitotic cells from non-mitotic cells. Cells with at least 1 associated spot were considered infected. Using the feature integration properties of CellClassifier, 6 population of nuclei (6 vectors) were calculated and exported to Excel and as outlined image: Infected mitotic cells (orange outline), non infected mitotic cells (pink, very rare therefore not shown), cell with mitotic neighbor, infected (green), mitotic neighbor, non-infected (cyan), normal cell, infected (red), non-infected (blue). Scale bar: 100 μm. B: Illustration of CellClassifier graphical output (heat maps). The experiment was done in a 96-well plate. Row A represents mock conditions (A01-A06) and no bacteria (A07-A12). The diagram in the upper left shows percentage of normal cells infected; the diagram in the upper right cell shows the log2 of the percentage of infected normal cells, normalized to a standard (G01-G12). Lower left: histogram of percent infected cells in the whole 96-well plate. Lower right: % normal cells infected plotted as a function of number nuclei. The red circles represent a trend analysis determined by a sliding window. In this plot, no gross trends are obvious. C: Summary of the output done outside CellClassifier, showing the preference of Salmonella docking for mitotic cells. D: Part of the Excel-file generated by CellClassifier.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2821321&req=5

Figure 4: Feature integration by CellClassifier. A: Salmonella (shown in green) preferentially dock onto mitotic cells (nuclei shown in grey). Segmentation and measurements of image were done using CellProfiler: First both, nuclei and Salmonella, were identified as independent objects. Cell objects were generated by expansion with reference to the nucleus. Inter-object relationships between Salmonella-spots and cells, as well as between neighboring nuclei were calculated. In CellClassifier a model was trained to distinguish mitotic cells from non-mitotic cells. Cells with at least 1 associated spot were considered infected. Using the feature integration properties of CellClassifier, 6 population of nuclei (6 vectors) were calculated and exported to Excel and as outlined image: Infected mitotic cells (orange outline), non infected mitotic cells (pink, very rare therefore not shown), cell with mitotic neighbor, infected (green), mitotic neighbor, non-infected (cyan), normal cell, infected (red), non-infected (blue). Scale bar: 100 μm. B: Illustration of CellClassifier graphical output (heat maps). The experiment was done in a 96-well plate. Row A represents mock conditions (A01-A06) and no bacteria (A07-A12). The diagram in the upper left shows percentage of normal cells infected; the diagram in the upper right cell shows the log2 of the percentage of infected normal cells, normalized to a standard (G01-G12). Lower left: histogram of percent infected cells in the whole 96-well plate. Lower right: % normal cells infected plotted as a function of number nuclei. The red circles represent a trend analysis determined by a sliding window. In this plot, no gross trends are obvious. C: Summary of the output done outside CellClassifier, showing the preference of Salmonella docking for mitotic cells. D: Part of the Excel-file generated by CellClassifier.

Mentions: Most important for the user is the summary of the whole experiment in a comprehensible and human readable format. Our program generates four different kinds of output data: outlined images, Excel-files, graphical summaries and a Matlab readable output. Outlined images visualize a vector or the result of the classification for a given image; if for instance the user wanted to visualize the vector "mitotic cells" using a yellow color, for all objects for which the vector had been positive (i.e. all mitotic cells) the outlines would be stained yellow (for examples see Figures 3, 4). Outlined images allow for a visual control of the final analysis and documentation. Excel-data are probably the most popular data format for biologists; all image, well and plate variables are automatically exported to an Excel-sheet. Well variables from the whole plate can be visualized as heat maps, histograms or scatter plots. They allow a quick overview over the whole experiment. When doing larger experiments, the user might want to do further customized analysis. Therefore, image and well variables of interest can be exported in a Matlab readable format.


Enhanced CellClassifier: a multi-class classification tool for microscopy images.

Misselwitz B, Strittmatter G, Periaswamy B, Schlumberger MC, Rout S, Horvath P, Kozak K, Hardt WD - BMC Bioinformatics (2010)

Feature integration by CellClassifier. A: Salmonella (shown in green) preferentially dock onto mitotic cells (nuclei shown in grey). Segmentation and measurements of image were done using CellProfiler: First both, nuclei and Salmonella, were identified as independent objects. Cell objects were generated by expansion with reference to the nucleus. Inter-object relationships between Salmonella-spots and cells, as well as between neighboring nuclei were calculated. In CellClassifier a model was trained to distinguish mitotic cells from non-mitotic cells. Cells with at least 1 associated spot were considered infected. Using the feature integration properties of CellClassifier, 6 population of nuclei (6 vectors) were calculated and exported to Excel and as outlined image: Infected mitotic cells (orange outline), non infected mitotic cells (pink, very rare therefore not shown), cell with mitotic neighbor, infected (green), mitotic neighbor, non-infected (cyan), normal cell, infected (red), non-infected (blue). Scale bar: 100 μm. B: Illustration of CellClassifier graphical output (heat maps). The experiment was done in a 96-well plate. Row A represents mock conditions (A01-A06) and no bacteria (A07-A12). The diagram in the upper left shows percentage of normal cells infected; the diagram in the upper right cell shows the log2 of the percentage of infected normal cells, normalized to a standard (G01-G12). Lower left: histogram of percent infected cells in the whole 96-well plate. Lower right: % normal cells infected plotted as a function of number nuclei. The red circles represent a trend analysis determined by a sliding window. In this plot, no gross trends are obvious. C: Summary of the output done outside CellClassifier, showing the preference of Salmonella docking for mitotic cells. D: Part of the Excel-file generated by CellClassifier.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2821321&req=5

Figure 4: Feature integration by CellClassifier. A: Salmonella (shown in green) preferentially dock onto mitotic cells (nuclei shown in grey). Segmentation and measurements of image were done using CellProfiler: First both, nuclei and Salmonella, were identified as independent objects. Cell objects were generated by expansion with reference to the nucleus. Inter-object relationships between Salmonella-spots and cells, as well as between neighboring nuclei were calculated. In CellClassifier a model was trained to distinguish mitotic cells from non-mitotic cells. Cells with at least 1 associated spot were considered infected. Using the feature integration properties of CellClassifier, 6 population of nuclei (6 vectors) were calculated and exported to Excel and as outlined image: Infected mitotic cells (orange outline), non infected mitotic cells (pink, very rare therefore not shown), cell with mitotic neighbor, infected (green), mitotic neighbor, non-infected (cyan), normal cell, infected (red), non-infected (blue). Scale bar: 100 μm. B: Illustration of CellClassifier graphical output (heat maps). The experiment was done in a 96-well plate. Row A represents mock conditions (A01-A06) and no bacteria (A07-A12). The diagram in the upper left shows percentage of normal cells infected; the diagram in the upper right cell shows the log2 of the percentage of infected normal cells, normalized to a standard (G01-G12). Lower left: histogram of percent infected cells in the whole 96-well plate. Lower right: % normal cells infected plotted as a function of number nuclei. The red circles represent a trend analysis determined by a sliding window. In this plot, no gross trends are obvious. C: Summary of the output done outside CellClassifier, showing the preference of Salmonella docking for mitotic cells. D: Part of the Excel-file generated by CellClassifier.
Mentions: Most important for the user is the summary of the whole experiment in a comprehensible and human readable format. Our program generates four different kinds of output data: outlined images, Excel-files, graphical summaries and a Matlab readable output. Outlined images visualize a vector or the result of the classification for a given image; if for instance the user wanted to visualize the vector "mitotic cells" using a yellow color, for all objects for which the vector had been positive (i.e. all mitotic cells) the outlines would be stained yellow (for examples see Figures 3, 4). Outlined images allow for a visual control of the final analysis and documentation. Excel-data are probably the most popular data format for biologists; all image, well and plate variables are automatically exported to an Excel-sheet. Well variables from the whole plate can be visualized as heat maps, histograms or scatter plots. They allow a quick overview over the whole experiment. When doing larger experiments, the user might want to do further customized analysis. Therefore, image and well variables of interest can be exported in a Matlab readable format.

Bottom Line: This represents a significant obstacle in many biology laboratories.Classification results can be integrated with other object measurements including inter-object relationships.This should facilitate the implementation of automated high-content screening.

View Article: PubMed Central - HTML - PubMed

Affiliation: Institute of Microbiology, ETH Zurich, Zürich, Switzerland. misselwitz@micro.biol.ethz.ch

ABSTRACT

Background: Light microscopy is of central importance in cell biology. The recent introduction of automated high content screening has expanded this technology towards automation of experiments and performing large scale perturbation assays. Nevertheless, evaluation of microscopy data continues to be a bottleneck in many projects. Currently, among open source software, CellProfiler and its extension Analyst are widely used in automated image processing. Even though revolutionizing image analysis in current biology, some routine and many advanced tasks are either not supported or require programming skills of the researcher. This represents a significant obstacle in many biology laboratories.

Results: We have developed a tool, Enhanced CellClassifier, which circumvents this obstacle. Enhanced CellClassifier starts from images analyzed by CellProfiler, and allows multi-class classification using a Support Vector Machine algorithm. Training of objects can be done by clicking directly "on the microscopy image" in several intuitive training modes. Many routine tasks like out-of focus exclusion and well summary are also supported. Classification results can be integrated with other object measurements including inter-object relationships. This makes a detailed interpretation of the image possible, allowing the differentiation of many complex phenotypes. For the generation of the output, image, well and plate data are dynamically extracted and summarized. The output can be generated as graphs, Excel-files, images with projections of the final analysis and exported as variables.

Conclusion: Here we describe Enhanced CellClassifier which allows multiple class classification, elucidating complex phenotypes. Our tool is designed for the biologist who wants both, simple and flexible analysis of images without requiring programming skills. This should facilitate the implementation of automated high-content screening.

Show MeSH
Related in: MedlinePlus